This is a library for using the Mapper for temporal topic modelling. Though things broadly work now, the edge cases have not been throughly tested.
Direct questions to Kaleb D. Ruscitti: kaleb.ruscitti at uwaterloo.ca .
Complete documentation is under construction on Read The Docs.
From the arXiv API, we can retrieve ~500,000 article titles and abstracts,
use SBERT to embed them, and then UMAP to reduce to 2D.
Using DataMapPlot and TopicNaming we can produce a static plot of this data:
Now, using this repository we can additionally analyse the temporal information. Using the Mapper algorithm with time as our lens function, we create a temporal graph of the topics (clusters) through time. The code includes two types of plots to visualize this graph:
| Centroid Plot | Temporal-Semantic Plot |
|---|---|
 |
 |
Install from PyPI:
pip install temporal-mapper
Or, clone the repo and install:
git clone https://github.com/TutteInstitute/temporal-mapper.git
cd temporal-mapper && pip install .
(Mostly for my future self...)
Clone the repo:
git clone https://github.com/TutteInstitute/temporal-mapper.git
Then make a virtual environment and install the package and pytest.
cd temporal-mapper && python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt && pip install -e .
Before making any changes, check that the tests run successfully:
cd tests && python -m pytest mapper.py